WP-01 / whose-instructions-count.md
Whose Instructions Count
The boundary model. Five questions — instructions, access, sharing, approval, and memory — each with the failure it prevents and the control that belongs beside it.
PAPERS / ARCHIVE
Two briefs. Room for the next one.
Both papers are written for holders, not as a certification. They restate the same problem in two forms: a control model you can check, and a doctrine you can hand to someone building the agent. Further briefs can be added here without changing the console.
WP-02 / why-a-security-policy.md
Why an Agentic NFT Needs a Security Policy
The doctrine. Personality does not set permissions. Outside content can hide commands. Flocks, tools, memory, teams, and token transfers each need an explicit line.